Plox
AI3 mins read

McKinsey Warns AI Agents Could Make Company AI Costs Harder to Predict

McKinsey says wider adoption of AI agents could push corporate AI spending higher and make costs more volatile, especially as agents complete multistep tasks in different ways.

Business Insider

Business Insider

McKinsey warns that AI may get a lot more expensive for companies

AI agent

AI Agents Bring a New Cost Problem

McKinsey & Company warns that companies’ AI spending may rise further as AI agents become more common. Unlike text-based AI tools, agents are designed to complete tasks, often through multistep processes that can take different routes to reach the same goal. That variability can make budgeting harder because the cost of completion can shift significantly from one agent or workflow to another.

The Big Number: Costs Can Vary by 30 Times

According to a McKinsey study cited in the article, the cost of completion for different agents could differ by as much as 30 times. Lari Hämäläinen, a McKinsey senior partner, compared the issue to running an operation where costs can swing by 30x from day to day. For business leaders, the takeaway is clear: AI pilots need cost controls and measurement before they scale.

AI Spending Is Already Becoming Material

McKinsey’s 2026 State of AI survey found that about a third of organizations spend more than 10% of their technology and communications budgets on AI. McKinsey said 60% of respondents planned to increase AI spending next year, while about one in five said AI spending was beginning to constrain operating costs. Tanguy Catlin, a McKinsey senior partner and director of the McKinsey Global Institute, said the spending is becoming “material and visible.”

Where Companies Are Feeling the Pressure

McKinsey said the issue is especially acute for software-development teams using agents to automate coding, which it described as “very token hungry.” The article notes that Amazon shut down an employee-created leaderboard tracking AI token use after some workers performed tasks to climb the rankings. Companies including Coinbase and Salesforce have also begun putting limits on AI use as bills rise.

Measure Value Before Scaling Agents

McKinsey has also said agentic AI can reduce the time humans spend on certain transformation-office tasks by 35% to 40%, and sometimes by 70% or more. But the firm argued that companies need to measure whether agents produce enough value to justify their cost. The practical implication: track both productivity gains and usage costs before making AI agents a default part of operations.

Discover More